A method of predicting speech intelligibility using computer algorithms is presented. Diagnostic Rhyme test number four was used to measure speech intelligibility using a subjective listener test and these results were used as a basis for comparison with the intelligibility predictions made by the computer algorithm. An audio recording of a speaker reading the Diagnostic Rhyme test was made. This recording was run through a General Electric radio system and varying amounts of noise were added. The output of the radio system was recorded, providing a copy of the input word corrupted by both additive noise and radio system distortion effects. Both the input recording and the noisy output recording were digitized by sampling the analog waveforms at a 10 kilohertz rate. These digital samples were converted to a frequency format by windowing the time samples with a rectangular window 128 time samples in length and processing them using Fast Fourier transform techniques.

plus-circle Add Review

comment

Reviews